Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | 강윤철 | - |
| dc.contributor.author | Angelica, Audrey | - |
| dc.creator | Angelica, Audrey | - |
| dc.date.accessioned | 2025-02-18T16:32:10Z | - |
| dc.date.available | 2025-02-18T16:32:10Z | - |
| dc.date.issued | 2025 | - |
| dc.identifier.other | OAK-000000243104 | - |
| dc.identifier.uri | https://dcollection.ewha.ac.kr/common/orgView/000000243104 | en_US |
| dc.identifier.uri | https://dspace.ewha.ac.kr/handle/2015.oak/270923 | - |
| dc.description.abstract | Click-through rate (CTR) prediction is an essential task in online advertising. This paper introduces a novel method that leverages Deep Spatio-Temporal Neural Networks (DSTN) integrated with ensemble and reinforcement learning techniques. It utilizes distinct auxiliary ad data—contextual, clicked, and unclicked—to independently generate and then dynamically merge predictions for more accurate Click-Through Rate (CTR) prediction. The study employs Deep Deterministic Policy Gradient (DDPG) to dynamically adjust weights during the combination process, enhancing the model's ability to rapidly adapt to changes in ad relevance and maximize the effectiveness of digital advertising campaigns.;클릭률(CTR) 예측은 온라인 광고에서 중요한 과제이다. 본 논문은 Deep Spatio-Temporal Neural Networks (DSTN)에 앙상블 학습과 강화 학습 기법을 통합한 새로운 방법을 제안한다. 이 방법은 문맥 광고, 클릭된 광고, 클릭되지 않은 광고와 같은 다양한 보조 광고 데이터를 독립적으로 예측한 후 이를 동적으로 결합하여 CTR 예측의 정확성을 높인다. 또한, 결합 과정에서 Deep Deterministic Policy Gradient (DDPG)를 사용해 가중치를 동적으로 조정함으로써 광고 관련성의 변화에 신속하게 적응하고 디지털 광고 캠페인의 효과를 극대화하는 모델의 능력을 향상시킨다. | - |
| dc.description.tableofcontents | I. Introduction 1 II. Related Work 4 A. Exploiting auxiliary data for CTR prediction 4 B. Limitations of DSTN and preceding CTR prediction models 5 C. Ensemble method for CTR prediction 5 D. Reinforcement Learning for CTR Prediction 6 E. Paper proposition 8 III. Methodology 9 A. Data Embedding and Processing 10 B. Deep Spatio-Temporal Neural Networks 12 C. Deep Deterministic Policy Gradient 16 D. Proposed model 18 IV. Experimental Settings 25 A. Research Questions and Method 25 B. Dataset 26 C. Compared Methods 27 D. Parameter Settings 30 V. Results 32 A. Training Results 32 B. Test Result 34 C. Research Questions Response 38 VI. Conclusion 40 Bibliography 42 Abstract(inKorean) 46 | - |
| dc.format | application/pdf | - |
| dc.format.extent | 1892516 bytes | - |
| dc.language | eng | - |
| dc.publisher | 이화여자대학교 대학원 | - |
| dc.subject | reinforcement learning, CTR, ensemble learning | - |
| dc.subject.ddc | 005.7 | - |
| dc.title | Enhancing User Click Probability Prediction Using Auxiliary Ads | - |
| dc.type | Master's Thesis | - |
| dc.title.subtitle | An Ensemble and Reinforcement Learning Approach | - |
| dc.format.page | iv, 46 p. | - |
| dc.identifier.thesisdegree | Master | - |
| dc.identifier.major | 대학원 빅데이터분석학협동과정 | - |
| dc.date.awarded | 2025. 2 | - |